Appropriate means to flexibly decide which operator to use in a particular situation at run time are not currently available
in generalization systems. These systems neglect important constraints imposed by cartographic principles which
may therefore lead to detrimental results of generalization solutions. The approach presented in this paper provides the
fundamentals for the development of comprehensive strategies for the generalization of categorical data. Based on a set
of generic constraints representing cartographic principles governing the generalization process and their subsequent
formalization, tools are provided to control and steer the generalization process on all levels of observation as well as to
evaluate the results. However, until a fully operational constraint-based generalization system is available further research
is necessary. Besides the development of improved algorithms for the treatment of categorical data in vector and
raster format, problems of system coordination at run time, such as operator and algorithm selection and adequate prioritizing
of constraints, still remain partly unsolved.